1 | /*! |
---|
2 | \file |
---|
3 | \brief DataSource for experiments with realistic simulator of the PMSM model |
---|
4 | \author Vaclav Smidl. |
---|
5 | |
---|
6 | ----------------------------------- |
---|
7 | BDM++ - C++ library for Bayesian Decision Making under Uncertainty |
---|
8 | |
---|
9 | Using IT++ for numerical operations |
---|
10 | ----------------------------------- |
---|
11 | */ |
---|
12 | |
---|
13 | #include <stat/loggers.h> |
---|
14 | #include <estim/kalman.h> |
---|
15 | #include "simulator_zdenek/simulator.h" |
---|
16 | #include "pmsm.h" |
---|
17 | |
---|
18 | using namespace bdm; |
---|
19 | |
---|
20 | //! Simulator of PMSM machine with predefined profile on omega |
---|
21 | class pmsmDS : public DS |
---|
22 | { |
---|
23 | |
---|
24 | protected: |
---|
25 | //! indeces of logged variables |
---|
26 | int L_x, L_ou, L_oy, L_iu, L_optu; |
---|
27 | //! Setpoints of omega in timespans given by dt_prof |
---|
28 | vec profileWw; |
---|
29 | //! Setpoints of Mz in timespans given by dt_prof |
---|
30 | vec profileMz; |
---|
31 | //! time-step for profiles |
---|
32 | double dt_prof; |
---|
33 | //! Number of miliseconds per discrete time step |
---|
34 | int Dt; |
---|
35 | //! options for logging, - log predictions of 'true' voltage |
---|
36 | bool opt_modu; |
---|
37 | //! options for logging, - |
---|
38 | public: |
---|
39 | //! Constructor with fixed sampling period |
---|
40 | pmsmDS () |
---|
41 | { |
---|
42 | Dt=125; |
---|
43 | Drv=RV ( "{o_ua o_ub o_ia o_ib t_ua t_ub o_om o_th Mz }" ); |
---|
44 | } |
---|
45 | void set_parameters ( double Rs0, double Ls0, double Fmag0, double Bf0, double p0, double kp0, double J0, double Uc0, double DT0, double dt0 ) |
---|
46 | { |
---|
47 | pmsmsim_set_parameters ( Rs0, Ls0, Fmag0, Bf0, p0, kp0, J0, Uc0, DT0, dt0 ); |
---|
48 | } |
---|
49 | //! parse options: "modelu" => opt_modu=true; |
---|
50 | void set_options ( string &opt ) |
---|
51 | { |
---|
52 | opt_modu = ( opt.find ( "modelu" ) !=string::npos ); |
---|
53 | } |
---|
54 | void getdata ( vec &dt ) |
---|
55 | { |
---|
56 | dt.set_subvector(0,vec ( KalmanObs,6 )); |
---|
57 | dt(6)=x[2]; |
---|
58 | dt(7)=x[3]; |
---|
59 | dt(8)=x[8]; |
---|
60 | } |
---|
61 | void write ( vec &ut ) {} |
---|
62 | |
---|
63 | void step() |
---|
64 | { |
---|
65 | static int ind=0; |
---|
66 | static double dW; // increase of W |
---|
67 | static double Ww; // W |
---|
68 | static double Mz; // W |
---|
69 | if ( t>=dt_prof*ind ) |
---|
70 | { |
---|
71 | ind++; |
---|
72 | // check omega profile and set dW |
---|
73 | if ( ind<profileWw.length() ) |
---|
74 | { |
---|
75 | //linear increase |
---|
76 | if ( profileWw.length() ==1 ) |
---|
77 | { |
---|
78 | Ww=profileWw ( 0 ); |
---|
79 | dW=0.0; |
---|
80 | } |
---|
81 | else |
---|
82 | { |
---|
83 | dW = profileWw ( ind )-profileWw ( ind-1 ); |
---|
84 | dW *=125e-6/dt_prof; |
---|
85 | } |
---|
86 | } |
---|
87 | else |
---|
88 | { |
---|
89 | dW = 0; |
---|
90 | } |
---|
91 | // Check Mz profile and set Mz |
---|
92 | if ( ind<profileMz.length() ) |
---|
93 | { |
---|
94 | //sudden increase |
---|
95 | Mz = profileMz(ind); |
---|
96 | } |
---|
97 | else |
---|
98 | { |
---|
99 | Mz = 0; |
---|
100 | } |
---|
101 | } |
---|
102 | Ww += dW; |
---|
103 | //Simulate Dt seconds! |
---|
104 | for ( int i=0; i<Dt; i++ ) |
---|
105 | { |
---|
106 | pmsmsim_step ( Ww , Mz); |
---|
107 | } |
---|
108 | // for ( int i=0;i<Dt;i++ ) { pmsmsim_noreg_step ( Ww , Mz);} |
---|
109 | |
---|
110 | //discretization |
---|
111 | double ustep=1.2; |
---|
112 | KalmanObs [ 0 ] = ustep*itpp::round( KalmanObs [ 0 ]/ ustep) ; |
---|
113 | KalmanObs [ 1 ] = ustep*itpp::round(KalmanObs [ 1 ]/ ustep); |
---|
114 | double istep=0.085; |
---|
115 | KalmanObs [ 2 ] = istep*itpp::round( KalmanObs [ 2 ]/ istep) ; |
---|
116 | KalmanObs [ 3 ] = istep*itpp::round(KalmanObs [ 3 ]/ istep); |
---|
117 | |
---|
118 | }; |
---|
119 | |
---|
120 | void log_add ( logger &L ) |
---|
121 | { |
---|
122 | L_x = L.add ( rx, "x" ); |
---|
123 | L_oy = L.add ( ry, "o" ); |
---|
124 | L_ou = L.add ( ru, "o" ); |
---|
125 | L_iu = L.add ( ru, "t" ); |
---|
126 | // log differences |
---|
127 | if ( opt_modu ) |
---|
128 | { |
---|
129 | L_optu = L.add ( ru, "model" ); |
---|
130 | } |
---|
131 | } |
---|
132 | |
---|
133 | void logit ( logger &L ) |
---|
134 | { |
---|
135 | L.logit ( L_x, vec ( x,4 ) ); |
---|
136 | L.logit ( L_oy, vec_2 ( KalmanObs[2],KalmanObs[3] ) ); |
---|
137 | L.logit ( L_ou, vec_2 ( KalmanObs[0],KalmanObs[1] ) ); |
---|
138 | L.logit ( L_iu, vec_2 ( KalmanObs[4],KalmanObs[5] ) ); |
---|
139 | if ( opt_modu ) |
---|
140 | { |
---|
141 | double sq3=sqrt ( 3.0 ); |
---|
142 | double ua,ub; |
---|
143 | double i1=x[0]; |
---|
144 | double i2=0.5* ( -i1+sq3*x[1] ); |
---|
145 | double i3=0.5* ( -i1-sq3*x[1] ); |
---|
146 | double u1=KalmanObs[0]; |
---|
147 | double u2=0.5* ( -u1+sq3*KalmanObs[1] ); |
---|
148 | double u3=0.5* ( -u1-sq3*KalmanObs[1] ); |
---|
149 | |
---|
150 | double du1=1.4* ( double ( i1>0.3 ) - double ( i1<-0.3 ) ) +0.2*i1; |
---|
151 | double du2=1.4* ( double ( i2>0.3 ) - double ( i2<-0.3 ) ) +0.2*i2; |
---|
152 | double du3=1.4* ( double ( i3>0.3 ) - double ( i3<-0.3 ) ) +0.2*i3; |
---|
153 | ua = ( 2.0* ( u1-du1 )- ( u2-du2 )- ( u3-du3 ) ) /3.0; |
---|
154 | ub = ( ( u2-du2 )- ( u3-du3 ) ) /sq3; |
---|
155 | L.logit ( L_optu , vec_2 ( ua,ub ) ); |
---|
156 | } |
---|
157 | |
---|
158 | } |
---|
159 | |
---|
160 | void set_profile ( double dt, const vec &Ww, const vec &Mz ) |
---|
161 | { |
---|
162 | dt_prof=dt; |
---|
163 | profileWw=Ww; |
---|
164 | profileMz=Mz; |
---|
165 | } |
---|
166 | |
---|
167 | void from_setting( const Setting &root ) |
---|
168 | { |
---|
169 | UI::SettingResolver params_exp(root["params"]); |
---|
170 | const Setting& params=params_exp.result; |
---|
171 | |
---|
172 | set_parameters ( params["Rs"], params["Ls"], params["Fmag"], \ |
---|
173 | params["Bf"], params["p"], params["kp"], \ |
---|
174 | params["J"], params["Uc"], params["DT"], 1.0e-6 ); |
---|
175 | |
---|
176 | // Default values of profiles for omega and Mz |
---|
177 | vec profW=vec("1.0"); |
---|
178 | vec profM=vec("0.0"); |
---|
179 | double tstep=1.0; |
---|
180 | root.lookupValue( "tstep", tstep ); |
---|
181 | UI::get( profW, root, "profileW" ); |
---|
182 | UI::get( profM, root, "profileM" ); |
---|
183 | set_profile (tstep , profW, profM); |
---|
184 | |
---|
185 | string opts; |
---|
186 | if ( root.lookupValue( "options", opts ) ) |
---|
187 | set_options(opts); |
---|
188 | } |
---|
189 | |
---|
190 | // TODO dodelat void to_setting( Setting &root ) const; |
---|
191 | }; |
---|
192 | |
---|
193 | UIREGISTER ( pmsmDS ); |
---|
194 | |
---|
195 | |
---|
196 | //! This class behaves like BM but it is evaluating EKF |
---|
197 | class pmsmCRB : public EKFfull |
---|
198 | { |
---|
199 | protected: |
---|
200 | vec interr; |
---|
201 | vec old_true; |
---|
202 | vec secder; |
---|
203 | int L_CRB; |
---|
204 | int L_err; |
---|
205 | int L_sec; |
---|
206 | public: |
---|
207 | //! constructor |
---|
208 | pmsmCRB():EKFfull() |
---|
209 | { |
---|
210 | old_true=zeros(6); |
---|
211 | } |
---|
212 | |
---|
213 | void bayes(const vec &dt) |
---|
214 | { |
---|
215 | static vec umin(2); |
---|
216 | vec u(2); |
---|
217 | //assume we know state exactly: |
---|
218 | vec true_state=vec(x,4); // read from pmsm |
---|
219 | E.set_mu(true_state); |
---|
220 | mu=true_state; |
---|
221 | |
---|
222 | //integration error |
---|
223 | old_true(4)=KalmanObs[4]; |
---|
224 | old_true(5)=KalmanObs[5];// add U |
---|
225 | u(0) = KalmanObs[0]; // use the required value for derivatives |
---|
226 | u(1) = KalmanObs[1]; |
---|
227 | interr = (true_state - pfxu->eval(old_true)); |
---|
228 | |
---|
229 | //second derivative |
---|
230 | IMpmsm2o* pf = dynamic_cast<IMpmsm2o*>(pfxu); |
---|
231 | if (pf) |
---|
232 | { |
---|
233 | secder=pf->eval2o(u-umin); |
---|
234 | } |
---|
235 | |
---|
236 | umin =u; |
---|
237 | EKFfull::bayes(dt); |
---|
238 | old_true.set_subvector(0,true_state); |
---|
239 | } |
---|
240 | |
---|
241 | void log_add(logger &L, const string &name="" ) |
---|
242 | { |
---|
243 | L_CRB=L.add(rx,"crb"); |
---|
244 | L_err=L.add(rx,"err"); |
---|
245 | L_sec=L.add(rx,"d2"); |
---|
246 | } |
---|
247 | void logit(logger &L) |
---|
248 | { |
---|
249 | L.logit(L_err, interr); |
---|
250 | L.logit(L_CRB,diag(_R())); |
---|
251 | L.logit(L_sec,secder); |
---|
252 | } |
---|
253 | |
---|
254 | void from_setting( const Setting &root ) |
---|
255 | { |
---|
256 | diffbifn* IM = UI::build<diffbifn>(root, "IM"); |
---|
257 | diffbifn* OM = UI::build<diffbifn>(root, "OM"); |
---|
258 | |
---|
259 | //parameters |
---|
260 | |
---|
261 | //statistics |
---|
262 | int dim=IM->dimension(); |
---|
263 | |
---|
264 | vec mu0; |
---|
265 | if(root.exists("mu0")) |
---|
266 | UI::get( mu0, root, "mu0"); |
---|
267 | else |
---|
268 | mu0=zeros(dim); |
---|
269 | |
---|
270 | mat P0; |
---|
271 | if(root.exists( "dP0" )) |
---|
272 | { |
---|
273 | vec dP0; |
---|
274 | UI::get(dP0,root, "dP0"); |
---|
275 | P0=diag(dP0); |
---|
276 | } |
---|
277 | else if (root.exists("P0")) |
---|
278 | UI::get(P0,root, "P0"); |
---|
279 | else |
---|
280 | P0=eye(dim); |
---|
281 | |
---|
282 | set_statistics(mu0,P0); |
---|
283 | |
---|
284 | vec dQ; |
---|
285 | UI::get( dQ, root, "dQ"); |
---|
286 | vec dR; |
---|
287 | UI::get( dR, root, "dR"); |
---|
288 | set_parameters(IM, OM, diag(dQ) , diag(dR)); |
---|
289 | |
---|
290 | //connect |
---|
291 | RV* drv = UI::build<RV>(root, "drv"); |
---|
292 | set_drv(*drv); |
---|
293 | RV* rv = UI::build<RV>(root, "rv"); |
---|
294 | set_rv(*rv); |
---|
295 | } |
---|
296 | |
---|
297 | // TODO dodelat void to_setting( Setting &root ) const; |
---|
298 | }; |
---|
299 | |
---|
300 | UIREGISTER ( pmsmCRB ); |
---|
301 | |
---|
302 | |
---|
303 | //! This class behaves like BM but it is evaluating EKF |
---|
304 | class pmsmCRBMz : public EKFfull |
---|
305 | { |
---|
306 | protected: |
---|
307 | int L_CRB; |
---|
308 | public: |
---|
309 | //! constructor |
---|
310 | pmsmCRBMz():EKFfull() {} |
---|
311 | |
---|
312 | void bayes(const vec &dt) |
---|
313 | { |
---|
314 | //assume we know state exactly: |
---|
315 | vec true_state(5); |
---|
316 | true_state.set_subvector(0,vec(x,4)); // read from pmsm |
---|
317 | true_state(4)=x[8]; |
---|
318 | |
---|
319 | E.set_mu(true_state); |
---|
320 | mu = true_state; |
---|
321 | //hack for ut |
---|
322 | EKFfull::bayes(dt); |
---|
323 | } |
---|
324 | |
---|
325 | void log_add(logger &L, const string &name="" ) |
---|
326 | { |
---|
327 | L_CRB=L.add(concat(rx,RV("Mz",1,0)),"crbz"); |
---|
328 | } |
---|
329 | void logit(logger &L) |
---|
330 | { |
---|
331 | L.logit(L_CRB,diag(_R())); |
---|
332 | } |
---|
333 | |
---|
334 | void from_setting( const Setting &root ) |
---|
335 | { |
---|
336 | diffbifn* IM = UI::build<diffbifn>(root,"IM"); |
---|
337 | diffbifn* OM = UI::build<diffbifn>(root,"OM"); |
---|
338 | |
---|
339 | //statistics |
---|
340 | int dim=IM->dimension(); |
---|
341 | vec mu0; |
---|
342 | if( root.exists( "mu0")) |
---|
343 | UI::get(mu0, root, "mu0"); |
---|
344 | else |
---|
345 | mu0=zeros(dim); |
---|
346 | |
---|
347 | mat P0; |
---|
348 | |
---|
349 | if(root.exists("dP0")) |
---|
350 | { |
---|
351 | vec dP0; |
---|
352 | UI::get(dP0, root, "dP0"); |
---|
353 | P0=diag(dP0); |
---|
354 | } |
---|
355 | else if(root.exists("P0")) |
---|
356 | UI::get( P0, root, "P0" ); |
---|
357 | else |
---|
358 | P0=eye(dim); |
---|
359 | |
---|
360 | set_statistics(mu0,P0); |
---|
361 | |
---|
362 | vec dQ; |
---|
363 | UI::get(dQ, root, "dQ"); |
---|
364 | vec dR; |
---|
365 | UI::get(dR, root, "dR"); |
---|
366 | set_parameters(IM, OM, diag(dQ), diag(dR)); |
---|
367 | |
---|
368 | //connect |
---|
369 | RV* drv = UI::build<RV>(root, "drv"); |
---|
370 | set_drv(*drv); |
---|
371 | RV* rv = UI::build<RV>(root, "rv"); |
---|
372 | set_rv(*rv); |
---|
373 | } |
---|
374 | |
---|
375 | // TODO dodelat void to_setting( Setting &root ) const; |
---|
376 | }; |
---|
377 | |
---|
378 | UIREGISTER ( pmsmCRBMz ); |
---|